Abstracts

Rationale: When invasive monitoring is required in epilepsy surgery patients it is important that subdural grids are optimally placed. Essentially, the seizure onset zone needs complete coverage. However, the more grids placed the higher the complication risk. For planning non-invasive techniques are available: PET, MEG, SPECT, EEG-fMRI. All of these typically produce results as 3D volumes that indicate regions involved in epilepsy. The challenge for the clinician is to integrate all 3D data, often presented independently, into a single strategy for placing electrode grids on the 2D cortical surface that is accessible to the surgeon. Here we present a single framework for integrating 3D presurgical data on the 2D cortical surface. This allows the clinician to plan grid placement in a consistent way. It also allows quantitative evaluation of the performance of noninvasive methods when ECoG recorded in grid electrodes can be considered the gold standard. Methods: Retrospectively, for a number of patients 3D MRI, interictal MEG spikes, interictal FDG-PET, post-implantation CT and (inter)ictal ECoG were available. All modalities were co-registered using CURRY NeuroImaging suite. A triangular mesh (edge 3 mm) of the surface enveloping the cortex was set up. Based on this mesh a volume conductor model for MEG was defined, and interictal spikes were modeled using the MUSIC algorithm. Resulting 3D points (resolution 3mm) around the maximum of the MUSIC metric were stored. The PET hypometabolic areas were marked and enclosed 3D points (3mm) stored. Grid electrode positions were extracted from CT and projected on the surface mesh. Electrodes containing (inter)ictal activity in ECoG were marked. Subsequently, all data was processed in MATLAB. Pre-implantation processing steps included: (1) determination of direction of center-of-mass (CM) of 3D points (PET/MEG) to nearest point on the surface mesh, (2) projection on the mesh of 3D points along this direction and marking of triangles involved, (3) computation of distance between projected CM's and overlap between marked triangles for PET/MEG. All results were plotted on the surface mesh and presented to the clinician with the computed measures. Post-implantation steps were: (1) triangles that represent sampling areas of grid electrodes containing (inter)ictal ECoG activity are marked and CM is computed, (2) calculation of distance and overlap between marked areas for CM PET/MEG and CM ECoG. Results: Results indicate that for some patients grid placement strategy would have changed if pre-implantation results were presented according to the framework introduced here. A subfrontal grid would, e.g., now be favored over a temporal pole position based on MEG results which were initially regarded as ambiguous. Post-implantation results show that interictal PET and MEG can indicate separate regions with different distances and overlap with ECoG. Conclusions: The framework presented here supports decision making in subdural grid placement. The proposed method allows quantification of the relation between PET, MEG, and (inter)ictal ECoG in future studies.